Stochastic Simulation

G7 allows models to be run with stochastic simulation. To do stochastic simulation with a model, one or more of its regression equations must be estimated after giving G7 the command

stochastic y

In the save (.SAV) files, the equation then will be followed by the letter “s”, the Standard Error of Estimate of the equation (SEE), the value of Rho, and the variance-covariance matrix of the regression coefficients. Running Build with such save (.SAV) files automatically will create a model that can be run with stochastic simulation. To run stochastic simulations, one must select the stochastic option in the run-model window and specify the following additional options:

Number of iterations:

This is the number of model solutions for random simulation; the default is 50.

Additive errors:

Check this box to run the simulation with random additive errors.

Random Coefficients:

Check this box to run the simulation with random coefficients.

In each iteration, values of the additive errors, or of the random coefficients, or both are picked and a deterministic simulation is run with those values. The additive errors will have the standard errors and autocorrelation coefficient found from the residuals in the regression. The random coefficients will have the variance-covariance matrix variance-covariance matrix found for the regression coefficients in the regression.

A run with stochastic simulation creates two banks. One is the usual output bank specified on the form which appeared when you clicked Model | Run. In the case of stochastic simulation, this bank contains the means of the various simulations. The second bank created by the stochastic simulation always is called SIGMA. It contains the standard errors of the simulations of the various variables in the model.